subaligner.hyperparameters module

class subaligner.hyperparameters.Hyperparameters

Bases: object

The configuration on hyperparameters used for training

OPTIMIZERS = ['adadelta', 'adagrad', 'adam', 'adamax', 'ftrl', 'nadam', 'rmsprop', 'sgd']

property back_hidden_size

property batch_size

clone() → subaligner.hyperparameters.Hyperparameters

Make a cloned hyperparameters object

Returns

Hyperparameters – The cloned Hyperparameters object.

property dropout

property epochs

property es_min_delta

property es_mode

property es_patience

classmethod from_file(file_path: str) → subaligner.hyperparameters.Hyperparameters

Deserialise a file content into a Hyperparameters object

Parameters

{string} -- The path to the file containing hyperparameters. (file_path) –

Returns

Hyperparameters – The deserialised Hyperparameters object.

classmethod from_json(json_str: str) → subaligner.hyperparameters.Hyperparameters

Deserialise JSON string into a Hyperparameters object

Parameters

{string} -- Hyperparameters in JSON. (json_str) –

Returns

Hyperparameters – The deserialised Hyperparameters object.

property front_hidden_size

property learning_rate

property loss

property metrics

property monitor

property network_type

property optimizer

to_file(file_path: str) → None

Serialise hyperparameters into JSON and save the content to a file

Parameters

{string} -- The path to the file containing saved hyperparameters. (file_path) –

to_json() → str

Serialise hyperparameters into JSON string

Returns

string – The serialised hyperparameters in JSON

property validation_split